EQ plugin sound quality difference at 96KHz vs 44.1KHz ?

[kylen]

Well - I can tell when I can't see the trees for the forest so I thought I'd better ask...

I just began working at 96KHz again today just to see if I have the firepower. I'm remastering some 2-track mixes and I just can't find an EQ that is sounding like I want, well I found one it's PSP MasterQ (in FAT mode). OK fine, I work for a few hours but something is wrong in the back of my mind. I mean how come I only have 1 EQ that will sound good all of the sudden...

Now the obvious hits me - is it because I am working on a native 96KHz/32bit stereo file in Sonar5 all of the sudden? I'm going to go back to 44.1KHz and see if I get the same impression back there at the lower sample rate. But then I'm wondering if I'm just sinking to the lower quality level and anything sounds good...

Anyone else notice this kind of thing on 96KHz native wav files? I've got some pretty good EQs I think - they all can't be rotten all of the sudden so I'm thinking the 96KHz must be wierd when using the plugins.

Any idea or comments or similar discoveries at 96KHz? Thanks.

[Aleksey Vaneev]

First of all, make sure you audio card sounds as desirable at 96 kHz as it does at 44.1kHz. You may also perceive your audio card as a plug-in which applies its own coloration before the sound goes to speakers. At 44.1kHz it may apply one coloration over plug-ins you have inserted, while at 96kHz this coloration may become different, and so the whole plug-in chain is going to sound subtly different.

But of course, there are differences in sound between 96 kHz and 44.1kHz... And I think I'll put such warning in the future, at least for some Voxengo plug-ins. It is not always possible to get identical performance at different sample rates without using heavy processing algorithms (for example, use FIR filters instead of using efficient IIR filters).

On the other hand, Voxengo PHA-979 was optimized to run equally good at 44.1kHz and 96kHz - some compressors as well (due to their low dependence on Nyquist frequency limit). EQs are generally not equal at different sample rates, especially at higher frequencies. If you are equalizing lower frequencies only, EQ's performance should be identical, though.

Polyphase oversampling filtering used in some Voxengo plug-ins (those with 'High Quality' switch) should sound differently at 44.1kHz and 96kHz even theoretically. And this can't be avoided without using FIR filtering that can be easily made invariant to sample rate used.

[kylen]

Thanks for your response Aleksey. Good point about the audio card or D/A - I'm using a Benchmark DAC-1 and I think they upsample everything to 192KHz so there's another monkey wrench in the signal path.

I guess what I'll do for now is just record my 2-track stuff at 96KHz then use R8Brain Pro to downsample to 44.1KHz (Minimum Phase mode). Since this stuff is mostly for CD anyway the end user won't hear all the cool 96KHz sounds anyway. In addition my rig (P4 2.6GHz) pretty much chokes with a few mastering plugins at 96KHz. Periodically I test to see how far I can get - which is what I did today...

You're exactly right - the differences I noticed were in the highs - and it was pretty huge IMO. Now that you mention it I think I've seen this discussion before but didn't connect for some reason...

Back to 44.1KHz !

[juha_p]

Thanks for your response Aleksey. Good point about the audio card or D/A - I'm using a Benchmark DAC-1 and I think they upsample everything to 192KHz so there's another monkey wrench in the signal path.

I guess what I'll do for now is just record my 2-track stuff at 96KHz then use R8Brain Pro to downsample to 44.1KHz (Minimum Phase mode). Since this stuff is mostly for CD anyway the end user won't hear all the cool 96KHz sounds anyway. In addition my rig (P4 2.6GHz) pretty much chokes with a few mastering plugins at 96KHz. Periodically I test to see how far I can get - which is what I did today...

You're exactly right - the differences I noticed were in the highs - and it was pretty huge IMO. Now that you mention it I think I've seen this discussion before but didn't connect for some reason...

Back to 44.1KHz !

Actually, there maybe are no advantages to record @ 96 kHz when you are using 24 or 32-bit resolution and your goal is CD quality (16-bit/44.1 kHz), if your recording audio device (which you didn't mention yet) is good enough (aren't the DAC1 for output only?).

You always loose something when doing SRC from 96 kHz to 44.1 kHz (and maybe little even when dither from 24-bit to 16-bit).

Some SRC comparisons:
- http://src.infinitewave.ca/
- http://xserve1.bias-inc.com:16080/produ ... esampling/ (white paper as PDF)

Juha

[Left Headphone]

96k is just to hard on the cpu in most systems. Maybe Quad core or something will be a solution? But then your host will have to be able to use all of it...

[BertKoor]

Or try working in 88kHz instead of 96kHz, since the downsampling is much more transparent.

[hibidy]

As luck would have it.........all three sound cards I have sound different with diff sample rates.............aside from how diffenent they sound from each other!

The "guitarport/podxt" stuff (my lowest end) sounds ok at 44 or 48, hard to tell the difference.........but that is to be expected

the maudio audiophile 192 (pc) is really really REALLY picky, if not at 96 or higher.........sounds awful..........I mean, super big-deal difference.....!

the motu traveler is not so picky, but there is a "noticable" difference between 44.1 and 96..........but not so much with like 88.2 and 96.

Now, one thing to keep in mind......even though the cpu hit is higher with let's say 24/96.......the latency is AUTOMATICALLY better........even at buffer sizes of 512 or even 1024! and you get the mo' lowend (and the high end, though many cannot hear that!)

[kylen]

Actually, there maybe are no advantages to record @ 96 kHz when you are using 24 or 32-bit resolution and your goal is CD quality (16-bit/44.1 kHz), if your recording audio device (which you didn't mention yet) is good enough (aren't the DAC1 for output only?).

Oops I thought we were talking about monitoring, for record I use a Mytek Stereo96 A/D which sounds very noticably better at 96KHz than 44.1KHz. It's not a little bit better but a lot better. But you also mention something important - my goal. Currently it is CD but in addition I wanted to at least offer people an archive copy (for future remastering) at the highest best sounding native resolution possible.

It's just that I can't really work in 96KHz at the moment I suppose. This is my 3rd shot at it, the other 2 times my cpu hardware just couldn't do it. Now my cpu hardware can do it (barely) but the sound of my audio app & plugins is so different it's a little like starting over. Some of the mastering plugins I have just don't even sound good at 96KHz. So I'll transcribe at 96KHz for archival purposes, then do SRC to 44.1KHz.

You always loose something when doing SRC from 96 kHz to 44.1 kHz (and maybe little even when dither from 24-bit to 16-bit).

Some SRC comparisons:
- http://src.infinitewave.ca/
- http://xserve1.bias-inc.com:16080/produ ... esampling/ (white paper as PDF)
Juha

OK thanks for the reading I'll check it out. Right now I'm using Voxengo R8Brain Pro in minimum phase mode as SRC. I'll have to compare that against simply selecting 44.1KHz at the A/D. For me, when I A/B something, if I decide I have to do a blind test to figure it out (if it's that close) then there is really no advantage to the extra processing or additional step. So in cases where I don't need to archive at 96KHz I might just set the Mytek to 44.1KHz and transcribe.

[kylen]

96k is just to hard on the cpu in most systems. Maybe Quad core or something will be a solution? But then your host will have to be able to use all of it...

Computers are gettin heftier lately - when I get some bread I'll have to pump mine up - true! Even the burgers are gettin more firepower - they put 3 of those babies on one bun now...

[kylen]

Or try working in 88kHz instead of 96kHz, since the downsampling is much more transparent.

I'll have to try that and give it a listening test myself - I've heard varied experiences here. Math vs Listening, that kind of thing. I don't know how much of a difference it would make to R8Brain Pro sound wise. Good idea - I've got some tests to run!

[kylen]

Now, one thing to keep in mind......even though the cpu hit is higher with let's say 24/96.......the latency is AUTOMATICALLY better........even at buffer sizes of 512 or even 1024! and you get the mo' lowend (and the high end, though many cannot hear that!)

That's good to know about latency. And the sound is quite different. In the old Math vs Hearing Range vs Actual Sound argument personally I can't hear up to 30KHz myself using test tones but the quality and even the freaking air around your head just seems clearer at 96KHz. Not by a small amout - like "I'm squinting and think I can hear something", it's obvious like "oh great - now I have to redo all my 44.1KHz stuff".

But all is not lost at 44.1KHz - thanks to our ears and aural memory or whatever it is...once you're in the company of 44.1KHz sounds for a minute or so any memory of 96KHz vanishes and you're happy again.

[Rangtangtang]

Now, one thing to keep in mind......even though the cpu hit is higher with let's say 24/96.......the latency is AUTOMATICALLY better........even at buffer sizes of 512 or even 1024! and you get the mo' lowend (and the high end, though many cannot hear that!)

That's good to know about latency. And the sound is quite different. In the old Math vs Hearing Range vs Actual Sound argument personally I can't hear up to 30KHz myself using test tones but the quality and even the freaking air around your head just seems clearer at 96KHz. Not by a small amout - like "I'm squinting and think I can hear something", it's obvious like "oh great - now I have to redo all my 44.1KHz stuff".

But all is not lost at 44.1KHz - thanks to our ears and aural memory or whatever it is...once you're in the company of 44.1KHz sounds for a minute or so any memory of 96KHz vanishes and you're happy again.

[BertKoor]

Kylen, did you try working at 48kHz instead of 44.1 or 96kHz? 48kHz is exactly 1/4 the base frequency of 192kHz so that may work with less artefacts.

The difference you notice may be due to diverse up- and downsampling stages: in the host, in plugins, in the soundcard... The more you can prevent that from happening the better.

I noticed with my soundcard (it's not a soundblaster!) that it performs more neutral at 48kHz than at 44.1kHz. Not much, but it's measurable. I can't hear it, but other people may. I suppose it doesn't really change its clock speed, but internally sticks with 192kHz.

The "ideal" soundcard should have a chrystal running at 705.6kHz (or a multiple) so both 44.1 and 48kHz (and all their multiples) can be derived by either counting 160 ticks or 147 ticks. Dunno weather that's done in practice a lot...

[kylen]

Same to you pal!

[kylen]

Kylen, did you try working at 48kHz instead of 44.1 or 96kHz? 48kHz is exactly 1/4 the base frequency of 192kHz so that may work with less artefacts.

The difference you notice may be due to diverse up- and downsampling stages: in the host, in plugins, in the soundcard... The more you can prevent that from happening the better.

Thanks - good idea I'll have to try that. There's somethin funny in most of the plugins so I've got a pretty good test case!